The AI boom is still a power story. Policy is still a speech. Cheap drones are still waiting on a budget.

Three Latent conditions, five Dormant. Power is the only one that got secondary company this week — still no filing. Lists and letters are not plants. Drones are not a line item. No butterfly. Capital untested.

U.S. cash markets are closed. We are not reading Friday's tape. We are asking what has to become true in the physical economy before the tape is forced to notice.

We are seeing a mismatch the market has not had to resolve yet.

The mismatch is this: AI demand is being financed as if electricity and interconnection were elastic. The system that delivers that power is not. Load rises faster than generation, transformers, switchgear, and queues can clear. Sites slip. Capex shows up first at utilities and grid equipment, not as a named "power trade" on the tape. That chain is why this desk exists. It is still Latent because we do not yet have the document that would force a state change.

Power is the only condition that even got company this week

Secondary coverage clustered around three related observations, one independence group each: U.S. data-center buildouts hitting grid limits; scarce electricity as a brake on the AI boom; storage and solid-state transformers as workarounds. That is consistent with the thesis. It is not a utility capex filing, not an interconnection-queue print, and not OEM backlog in a 10-Q.

Two facts can both be true. The Street already talks about "AI power." We still lack the primary that would move this from Latent to Emerging. Attention is not confirmation. An ETF wrapping the buildout is attention with a ticker on it — still not a queue.

What argues against us: efficiency, on-site generation, or delayed training capex could slow load without the grid ever becoming the binding constraint. Vendor TAM reports on transformers are not system evidence. If interconnection requests and equipment lead times ease without a demand-shock explanation, the thesis is wrong. Clock: weeks to two quarters for a named-load filing; quarters for capacity FID.

Lists are not plants

The second Latent condition is midstream. Processing concentration plus export-control architecture could change who can finance magnets, metals, and memory — but only if policy becomes contracts and plants.

This week's scan produced Entity List commentary, a 1260H list update, and a House letter urging a block on Chinese memory chips. Those are restatements of standing architecture. A letter is not a rule. A list update is not a plant FID. Rhetoric is not appropriated spending.

Substitution, recycling, or demand destruction can cap prices without a new Western mine. Announced projects fail at permitting, offtake, and cost. Kill path: controls waived at scale, or cancellations with no replacement offtakes. Clock: weeks to quarters for a new control or award; quarters to years for capacity.

Cheap drones have a battlefield. They do not yet have a line item.

The third Latent condition is industrial, not cinematic. Cheap uncrewed systems changed the cost-exchange where they are used. Doctrine and procurement have not yet shown obligated dollars. Until an RDT&E or production line exists as money, this is a hypothesis about capacity and sensors — not confirmation of a program.

Countermeasures can collapse the cost advantage. Budgets can stay in legacy platforms. Announcements are not obligations. Clock: weeks to quarters for a contract that is not a reprint.

What we actually tested

Five source families were touched. Every item arrived as Level C Google News RSS. The news warehouse returned no hits. Primary origin was not independently verified on any story. That is the ceiling of this edition. It is not a claim that origin nodes were watched.

Scientific coverage collapsed into one independence group: FDA pressure on ClinicalTrials.gov reporting, April–July reprints of the same campaign. Compliance theater is not a new therapeutic mechanism. Healthcare stays Dormant. We will not seed it on reprints.

Capital coverage was activist 13Ds in unrelated software and specialty names. That does not test theme-level ownership of grid equipment, midstream minerals, or defense production. Capital state: pre-capital / not yet tested against LongShort theme-level ownership.

Gold, consumer, distress, and wildcards: nothing earned a why. Dormant is a researched scan, not an empty page.

Downstream tape

Intelligence's latest close — fragile tech/AI leadership while macro containment holds — is what the session meant. It does not originate the power constraint. It is the kind of tape against which a still-Latent physical bottleneck can later capitalize, or fail to. We record that divergence. We do not promote this desk from the other.

Butterfly ledger

No Spark survives the gate. Consumer/fashion color appeared without cluster or geographic reproduction. Opening a Watch would be chasing a weak signal. The ledger is researched-empty on purpose.

Carrying forward: AI power and grid; critical minerals supply chains; defense autonomy and counter-UAS. All Latent. All still owed a primary.

If we are right — and if we are not

If the grid thesis is right, the next evidence is a utility, developer, or OEM print that names data-center load or equipment lead times.

If the minerals thesis is right, the next evidence is a control, award, or FID that is not last year's industrial-policy speech.

If the defense thesis is right, the next evidence is obligated dollars, not a feature on last year's drone narrative.

Silence inside those windows is evidence. Kill paths stay those written at birth. We will not draw a ticker chain while the objects are still Latent.